Epilepsy is a dynamical disease. Over the past decade, many seizure prediction and detection algorithms have been proposed to anticipate seizures, however detection delays are 5-15 seconds after the seizure onset, which is too late to be used as a signal for perturbations that may abort the paroxysm before it starts. Furthermore, improving the performance of these algorithms by increasing computational power may slow down closed loop monitoring systems and may be not suitable for real-time applications.